Seminar "Parsimonious Inference"

Abstract:

 

In this talk, I give the basic idea to bridge the gap between inferential schemes rooted in logic and in statistics. 

The theory, which is constructed around the notion of constraints on real-valued functions, can be extended to

logic thanks to classic T-norms. The inferential mechanism that I propose focuses on a parsimonious selection

of the constraints to favor simple and elegant explanations. I give a few preliminary examples to provide 

experimental evidence of the theory.